Background and Purpose The limited walking ability that follows a stroke restricts the patient’s independent mobility about the home and community, a significant social handicap. To improve the in-hospital prediction of functional outcome, the relationships between impairment, disability, and handicap were assessed with clinical measures in 147 stroke patients.
Methods The patients’ level of functional walking ability at home and in the community was assigned by expert clinicians to one of the six categories of a modified Hoffer Functional Ambulation scale at least 3 months after discharge. A 19-item questionnaire was further used to assess current customary mobility of the subjects. Functional muscle strength and proprioception were tested, and walking velocity was measured.
Results The significant indicators of impairment, upright motor control knee flexion and extension strength, differentiated household from community ambulators. The addition of velocity improved the functional prediction. Proprioception was clinically normal in all walkers. The validity of the criteria for the six levels of walking handicap was confirmed statistically. Stepwise discriminant analysis reduced the ambulation activities on the questionnaire from 19 to 7. Redefinition of the criteria for patient classification using the coefficients and constants of the seven critical functions improved the prediction of patient walking ability to 84%.
Conclusions The results of this study offer a quantitative method of relating the social disadvantage of stroke patients to the impairment and disability sustained. The measurement of therapeutic outcome in relation to the social advantage for the patient would allow more efficient standardization of treatment and services.
The hemiplegia that commonly follows a stroke reduces the patient’s ability to walk. As a result, the person’s independence in moving about the home or community may be significantly compromised. Often patients are incapable of returning to their premorbidity roles in society or may even require long-term care. To facilitate appropriate planning for the home and family, clinicians are challenged to predict at the time of hospital discharge the degree of social handicap that the patient will experience. Few guidelines are available, however, to assist the clinician in meeting this responsibility. Medical examinations assess impairment. Reflexes, passive mobility, strength, and sensation are the most frequently used measures for the hemiplegic limb. The relationship between these measures and walking ability has not been defined. Several investigators have suggested formal gait analysis to identify the joint motions and spatial-temporal parameters as indexes of disability.1 2 Gait velocity has been related to the severity of impairment, but this information has not been correlated with home and community independence.
In the rehabilitation setting, several systems have been developed to grade the quality of the patient’s degree of walking independence after a stroke.3 4 5 6 Also, patient training in managing curbs, stairs, and ramps, as well as level surfaces, is provided when appropriate to prepare for community mobility.7 8 None of these approaches, however, evaluate treatment in terms of walking performance in the home and community.9 Hoffer et al10 proposed a four-step scale of walking handicap (unable, physiological, household, and community) for use with children who have myelomeningocele. Clinicians have found this to be a useful classification, but the differentiating criteria have not been verified. Additionally, these categories are broad and do not reflect the variable demands that adults experience in the community.
The purpose of the present study was to create and verify the criteria for an expanded version of the Hoffer scale that would be applicable to adults and to determine the relationship between selected indicators of impairment, disability, and social handicap for patients having residual hemiplegia after stroke.
Subjects and Methods
Subjects included 147 patients (68 men, 79 women) impaired by cerebrovascular accidents of different etiologies. Average age was 55.5±12.2 years, and average weight was 70.9±13.8 kg. Left hemiplegia was present in 54% of patients. Testing occurred at least 3 months after the onset of stroke and at least 6 weeks after the initial poststroke discharge from the hospital. All patients could walk at least 6 m, and all gave their free and informed consent.
To define a relationship between handicap, disability, and impairment in stroke patients with hemiplegia, a series of assessments were designed by a group of experienced clinicians (a doctor and physical therapists with more than 10 years of experience in treating stroke patients). The combination of interviews and tests was conducted during a routine outpatient visit. Five clinical measures were used as described below.
Functional walking category. Categories were developed to provide a gross classification of the patient’s walking handicap. The original classification of Hoffer et al10 was modified by deleting the nonambulatory level and increasing the household and community levels of ambulation to a total of six categories (Table 1⇓). The expert clinician group developed the criteria for these six categories. Each patient in the study was assigned to one category according to these criteria.
Walking ability questionnaire. A questionnaire was constructed to provide a more detailed assessment of the patient’s social limitations resulting from decreased walking ability. The questionnaire (Fig 1⇓), administered by an experienced physical therapist, rated the patient’s current customary mobility in terms of 19 ambulatory activities commonly performed in the home (8) and community (11). Current customary mobility was defined as the patient’s self-reported ability to enter and leave the listed locations. Mobility was classified as “independent,” “supervised,” “assisted,” “wheelchair,” or “unable.” Programs were designed to convert the subjective assessment ratings to numerical grades (4, 3, 2, 1, and 0, respectively). An overall score was calculated. The range of scale was 0 to 76.
Stride characteristics. Stride characteristics were measured by a Footswitch Stride Analyzer.12 13 Bilateral footswitches containing compression-closing switches under the heel and forefoot were taped to the sole of the patient’s foot. Subjects walked across a 10-m walkway with the middle 6 m designated by photoelectric cells as the data collection area. From the foot-floor contact patterns, velocity, stride length, cadence, and swing and stance durations were calculated.
Upright motor control test. The upright motor control (UMC) test14 was used to assess functional muscle strength of the hemiplegic limb. Voluntary control of the hip, knee, and ankle of the involved lower extremity was tested with the subject standing (Fig 2⇓). Extension was graded on a three-point scale according to the patient’s ability to extend the knee from a flexed position while in single-limb stance. Completing the full range of extension was graded as strong 3; just supporting body weight on the flexed knee was a moderate grade 2; and inability to bear full weight on a flexed knee was graded as weak 1. Flexors were graded on a two-point scale: strong 3 and weak 1, based on the completion of a rapid change of flexion at each joint. The “moderate” grade 2 was omitted because according to Toman15 it provided no additional information to that of grade 1. Each patient’s hip, knee, and ankle scores were summed to provide an individual score with a maximum of 18. Intertester reliability of the UMC test was determined to be satisfactory at the time of its development (J. Montgomery, unpublished data, 1975).
Proprioception. Proprioception was tested with the patient sitting and eyes closed. The patient was asked to repeat, with the unaffected leg, movements of the hip, knee, and ankle as performed by the assessor with the affected leg. Performance was graded as 3, normal (accurate and prompt); 1, impaired (accurate but delayed); or 0, absent (wrong). The hip, knee, and ankle scores were summed to provide an individual patient score with a maximum of 9.
For statistical analysis of the data collected on each patient, Hoffer scale assignment (1 item), questionnaire responses (19 data items), stride characteristics (24 data items), joint sensation (3 data items), and motor control testing (6 data items) were entered into a computer-based (VAX 750) information-handling system (RS/1, Bolt, Beranek, and Newman). Statistical analysis of the data was performed in three stages using bmdp statistical software programs. The data were checked for normality of distribution with the Wilks’ Shapiro ‘W’ statistic. The proprioception and UMC scores were not normally distributed; therefore, nonparametric statistics were used with these data. The stride characteristics and data from the walking ability questionnaire, however, were normally distributed. Parametric statistics were considered appropriate for the questionnaire scores because the range of patient scores available extended from 0 to 76.
The first analytical approach consisted of ANOVA or the nonparametric counterpart, Kruskal-Wallis test for the patients’ scores in each test (questionnaire responses, stride characteristics, proprioception, and upright motor control), to determine whether traditional clinical data showed significant differences between the six functional walking categories (Table 1⇑). Post hoc testing included a Bonferroni correction factor for probability values to account for multiple comparisons.
Stepwise discriminant analysis techniques16 were used to identify which of the 19 items on the walking questionnaire contributed most to the differentiation between the functional walking categories (Table 1⇑). The data sample was divided into two groups using random selection by computer. Eighty percent of the patients (117 of 147) served as the main data set, and the questionnaire items that best discriminated between subjects in the six walking categories were identified for this group. To determine the error in classifying a new group of subjects, the remaining 20% of the subjects (30 patients) were classified with a cross-validation procedure using the variables and coefficients that had been identified as critical for the 117 patients in the initial analysis. The agreement between clinician grouping and computer-assisted grouping of the patients’ functional walking ability also was determined. The probability of misclassification was determined for each case by computing (1) the distance between the point representing an individual and the point representing the estimated population mean (Mahalanobis’ distance or D2) and (2) the posterior probability of belonging to each group.
The final phase in identifying the questionnaire items needed to classify walking handicap used the critical variables (final questionnaire items) and their coefficients from the discriminant analysis to redefine the criteria for the six functional ambulation categories. Items with the greatest difference in coefficients between two adjacent groups were identified as the strongest differentiators of those groups. The minimum and maximum scores of those items that caused a patient to be classified in the higher or lower of two adjacent groups were determined for each of the six walking categories. The descriptors of the functional walking categories were modified to reflect the threshold levels of performance in the critical items. All of the 147 patients were classified using the new criteria. Agreement of grouping using the modified descriptors with classification by the expert clinicians and the computer was determined.
Stepwise discriminant analysis also was used to select the best predictors of the patients’ functional walking category from age, sex, weight, proprioception, and upright motor control and stride characteristic scores. The discriminant analysis was repeated to determine whether the predictors could differentiate the broader categories of home-bound ambulators and community walkers after they failed to show acceptable separation of the six functional walking categories.
The expert clinician’s classification of the 147 patients assigned 26 or 27 patients to each of the six functional walking categories (Table 1⇑), except for a smaller group of physiological walkers. Scores on the walking ability questionnaire ranged from 11 to 57, of a possible total of 76. There was a significant difference (P<.05) between each of the six groups (Table 2⇓).
Among the initial 19 functional ambulation items from the walking questionnaire, seven were found by stepwise discriminant analysis to be sufficient to discriminate between the functional walking categories (Table 3⇓). These were (1) bathroom mobility, (2) bedroom mobility, (3) entering and exiting the home, (4) ascending and descending a curb, (5) movement through grocery stores, (6) movement through a shopping center (uncrowded), and (7) movement through a shopping center (crowded). The resulting classification system consisted of a matrix of coefficients representing the weighting factors for the individual functions and a constant for each walking category (Table 3⇓). The category with the highest total value (y) for the sum of the constant and the seven variables multiplied by their respective coefficients predicted the level of walking handicap for an individual patient.
Walking category for the 117 patients as assigned by the clinicians agreed with the classification of the patients by computer using the seven critical walking functions in 84% of the cases. The jackknife procedure, which is an unbiased estimate of the accuracy of group assignment, placed the 117 patients by eliminating each subject in turn from the coefficients. Using this procedure, group placement by computer agreed with the clinician placement in 75% of the cases. The cross-validation procedure classified the subset of 30 patients (not included in the calculation of the coefficients) with an agreement of 83% with the expert clinician’s walking group assignment.
Analysis of the differences in the coefficients of the seven critical questionnaire items between the adjacent walking categories resulted in a redefinition of the discriminating criteria (Table 4⇓). When the 147 patients were classified using the refined descriptors of the levels of walking categories, group placement agreed with the clinicians in 82% of the cases and with the computer classification in 90%.
Of the impairment measures, proprioception was not significantly different between any of the six functional walking groups. The mean proprioception score was within normal range for all groups.
Composite UMC scores (Table 2⇑) did not significantly differentiate the six groups. There was a significant difference (P<.05) demonstrated, however, between the community walkers as a group with a mean±SD score of 14±4 and household walkers with a score of 9±3. Just the knee flexion and knee extension scores on the UMC test were sufficient to predict home versus community walking ability with 78% accuracy and 78% agreement on cross-validation. If either knee flexion or extension score was 3 (grade strong), the predicted walking category would be community ambulation (Tables 5⇓ and 6⇓). If both muscle group scores were 1 or 2 (weak or moderate), household walking ability would be at the predicted level.
All of the stride characteristics, serving as measures of walking disability, increased for each higher level of the six functional ambulatory categories. The difference in velocity, however, showed the greatest statistical significance between groups (Table 2⇑). Mean velocities ranged from 6±2 m/min for physiological walkers to 48±11 m/min for community walkers (Table 2⇑). Velocity was significantly different between the three groups of community walkers (most, least, and unlimited) but was similar for the three categories of household walkers. The unlimited household walkers and the most limited community walkers also were not differentiated by their gait velocities.
Discriminant analysis identified velocity as the only impairment/disability variable to significantly predict placement in the six walking categories. Agreement with the clinicians’ scoring was only 44%. Age, sex, and weight also had no predictive ability of walking ability. The combination of velocity and knee extension control was found to contribute most to the gross differentiation between household walkers and community walkers, showing 87% agreement with the expert clinicians. The cross-validation of the subset of 30 patients had 78% agreement with the clinicians. For a patient with a strong grade of knee extension, a gait velocity of 16 m/min would predict community ambulation ability, whereas patients with moderate and weak knee scores would require the additional walking skills (ie, postural substitutions) that permitted velocities of at least 24 m/min and 32 m/min, respectively, to indicate the higher level of walking (Tables 7⇓ and 8⇓).
Velocity alone as a determinant of household versus community ambulation status had a slightly lower agreement with the clinician assessment at 85% and a cross-validation agreement of 77%. Patients with an average velocity of 25 m/min or more would be predicted to have the ability for community ambulation (Table 9⇓).
This investigation has produced a quantitative system of assessing a person’s customary level of walking ability at home and in the community. According to the World Health Organization (WHO) criteria, this is a system for assessing walking handicap.11 In addition, clinical criteria were identified that could facilitate the prediction of community walking ability before discharge.
A major difficulty in defining walking handicap was identifying the nature of the social disadvantage suffered as a result of the reduction in walking ability. Walking impairment or disability can interfere with an individual’s ability to participate in activities of daily living, even though walking does not in itself directly fulfill any of the fundamental functions of ordinary everyday activity such as sleeping, washing, dressing, eating, and working. In all human societies, however, these activities take place in specific locations. The extent to which walking limitation creates a social disadvantage depends on how much reduced mobility interferes with the individual’s ability to access and move around those locations where the fundamental activities of daily life take place. Measurement of this was the focus of the walking ability questionnaire. The initial list of 19 items was reduced to seven functions that were critical to differentiating the functional walking categories (Table 1⇑). Each category represents a different level of walking handicap. As such, it overcomes the difficulties encountered by Grimby and coworkers17 in applying the WHO mobility handicap scale to analysis of functional recovery in 76 stroke patients.
The series of functional ambulation tasks used to classify walking ability included the management of three variables: (1) changes in level and terrain irregularity, (2) obstacle avoidance, and (3) increase in distance. Mastery of a change in level is the critical task for entering and leaving the home and for curb management. Ability to manage stairs was not a critical task. Although this would appear to be a factor associated with the one-level housing in southern California, a pilot test of the system on 60 persons in Dublin confirmed the finding. Management of obstacles appears to be the critical functional ambulation task for shopping. The identification of shopping locations as a critical challenge for the community walker suggests that the highest level of functional walking ability incorporates a fourth variable: (4) manual handling of loads.
The results of this study (Table 10⇓) indicate that the limitation of customary mobility in normal everyday life seen in individuals after stroke is much higher than the figure of 20% given in earlier studies.18 Correlation of the seven critical functional ambulation activities with the six grades of function showed that assistance from another person was needed by all patients limited to the household. Also, the most limited community walkers had gained independence only in the management of irregular terrain (home exit and curb).
Clinical experience suggests that the two most important measures of impairment in determining walking ability are muscle strength and proprioception. Keenan et al14 have previously shown that upright motor control and proprioception were more strongly related to equilibrium than were motor planning, body awareness, muscle tone, or sensory integration. The present study confirmed these observations. Normal levels of proprioception were seen in all categories of walkers despite substantial variations in motor ability. The suggestion that proprioception is an essential precondition for independent walking at any level is consistent with the findings of Litchman et al.19
For the patients included in this study, knee extension and flexion were the discriminating elements in the UMC test (Fig 2⇑). Because the test is administered with the patient standing, the UMC tests are directed to the demands of walking (Fig 2⇑). In this way, the problems arising from limb posture, vestibular tone, patterns of muscle coordination, and length of muscle are included. Although it does not aid in the classification of the six categories, the knee UMC score proved useful in predicting household versus community walking ability before hospital discharge. The finding that only knee upright control was related to function is both convenient and logical. To maintain upright balance as one extends a previously flexed knee while bearing weight on just that limb would require postural changes at the ankle and hip as well. Hence, the quality of upright control at the knee may represent total limb-control capability, with the isolated hip and ankle tests being less significant and redundant fragments.
By combining the knee extension score and gait velocity, the in-hospital prediction potential was increased. This technique appears to reflect the influence of other factors such as spasticity and contracture. For example, 90% of the household walkers in this sample had knee extension scores of 1 or 2, yet the overlay of other limitations restricted the walking ability of a few patients with grade 3 knee extensor control. Similarly, 14% of the community ambulators had only grade 1 knee extensors. This reflects either their ability to substitute or nonvolitional stability from spasticity. The classification of patients with moderate (grade 2) knee extension was predominantly dependent on their gait velocities.
Velocity, as an independent measure, also showed a strong potential for differentiating the household and community ambulators. This finding supports the claims by others that velocity is the most critical of the stride characteristics.3 20 21 22 Velocity alone, however, was not a sufficiently strong predictor to serve as the sole classifier of ambulation levels, as has been suggested.23 Velocity did not increase significantly between the levels of household walkers. Claims of increased functional walking performance based on velocities less than 22 to 25 m/min must be viewed with caution.24 The mean velocity of the highest category of community walker was 48 m/min compared with the velocity of the normal, healthy population of 80 m/min.25 This finding supports goal setting based on the values typically displayed by functionally independent stroke patients22 rather than those of healthy subjects.26 This lower standard, however, represents a functional compromise: it is adequate for a stroke patient’s typical activities yet less than the normal ability needed to cross a wide commercial street within the traffic signal time.27
Establishing relationships between ability and handicap allows triaging of treatment goals. The patient is entitled to a reasonable prognosis as to functional walking ability. Overly optimistic expectations as well as underachievement must be avoided. There is some evidence that in stroke patients the least able receive the most rehabilitation and the initially more able receive least.28 Keenan and coworkers14 showed that only 50% of surviving stroke patients manage to walk in the community. This finding was corroborated in the present study. The possibility that the initially more able patient will not achieve his or her full potential must be considered. The findings of this study offer criteria for determining the optimal training programs in mobility without unreasonable prolongation or denial of treatment. Intact proprioception is a basic requirement for independence. For walkers with community potential, the additional determinants are the UMC knee score, gait velocity, and the ability to manage changes in level (Table 3⇑). This confers access to the world outside the walls of the home. Tolerance of crowded environments and some manual handling skills (eg, carrying a newspaper or shopping bag) should allow further advance in the level of community walking together with an increase in walking velocity. For household walkers, the ability to move around and to access the bedroom and bathroom constitute an advance in independence for the patient and relief for the caregiver.
The results of this investigation should provide a means to improve communication about the handicaps resulting from stroke in terms of patient mobility. Treatment and services to maximize the patient’s social advantage can become more standardized. In addition, this study provides a means of conveying to the patient a realistic expectation of treatment outcome regarding his or her customary mobility in normal everyday life.
This study was supported by grant G008300077 from the National Institute of Disability and Rehabilitation Research. A European Commission contribution in the framework of the Advanced Informatics in Medicine program CAMARC II project was received to defray travel expenses involved in the preparation of manuscripts. Florence Grehan, Photographic Department, Mater Hospital, Dublin, Ireland, provided the photographs.
Reprint requests to Dr M. Garrett, University College Dublin School of Physiotherapy, Eccles St, Dublin 7, Ireland.
- Received August 25, 1994.
- Revision received March 1, 1995.
- Accepted March 1, 1995.
- Copyright © 1995 by American Heart Association
Saleh M, Murdoch G. In defence of gait analysis. J Bone Joint Surg Br. 1985;67:237-241.
Holden MK, Gill KM, Magliozzi MR, Nathan J, Piehl-Baker L. Gait assessment for the neurologically impaired: reliability and meaningfulness. Phys Ther. 1984;64:35-40.
Gowland C, Stratford P, Moreland J, Torresin W, Van Hullenar S, Sanford J, Barreca S, Vanspall B, Plews N. Measuring physical impairment and disability with the Chedoke-McMaster Stroke Assessment. Stroke. 1993;24:58-63.
Anderson TP. Rehabilitation of patients with completed strokes. In: Kottke FJ, Lehmann JF, eds. Krusen’s Handbook of Physical Medicine and Rehabilitation. 4th ed. Philadelphia, Pa: WB Saunders Co; 1990:656-678.
Leslie LR. Training for functional independence. In: Kottke FJ, Lehmann JF, eds. Krusen’s Handbook of Physical Medicine and Rehabilitation. 4th ed. Philadelphia, Pa: WB Saunders Co; 1990:564-570.
Thorngren M, Westling B, Norrving B. Outcome after stroke in patients discharged to independent living. Stroke. 1990;21:236-240.
Hoffer MM, Feiwell E, Perry R, Perry J, Bonnett C. Functional ambulation in patients with myelomeningocele. J Bone Joint Surg Am. 1973;5:137-148.
World Health Organization. International Classification of Impairments, Disabilities, and Handicaps. Geneva, Switzerland: World Health Organization; 1980.
Perry J. Clinical gait analyzer. Bull Prosthet Res. 1974;Fall:188-192.
Bontrager E. Footswitch Stride Analyzer. Bull Prosthet Res. 1981;18:284-288.
Keenan MA, Perry J, Jordan C. Factors affecting balance and ambulation following stroke. Clin Orthop. 1984;182:165-171.
Toman CJ. Upright control as a predictive gait measure in patients with hemiplegia. Los Angeles, Calif: University of Southern California; 1983. Dissertation.
Afifi AA, Azen SP. Statistical Analysis, a Computer-Oriented Approach. 2nd ed. New York, NY: Academic Press Inc; 1979:310-318.
Gresham GE, Fitzpatrick TE, Wolf PA, McNamara PM, Kannel WB, Dawber TR. Residual disability in survivors of stroke: the Framingham Study. N Engl J Med. 1975;19:954-956.
Litchman CD, Dietz M, Reding MJ. Clinical examination is more predictive of ambulation recovery than CT lesion location following subcortical ischemic stroke. J Neurol Rehabil. 1994;8:55-61.
Holden MK, Gill KM, Magliozzi MR. Gait assessment for neurologically impaired: standards for outcome assessment. Phys Ther. 1986;66:1530-1539.
Perry J. Gait Analysis: Normal and Pathological Function. Thorofare, NJ: SLACK Inc; 1992:431-433.
Lerner-Frankiel MB, Varcas S, Brown MB, Krusel L, Schoneberger W. Functional community ambulation: what are your criteria? Clin Man Phys Ther. 1986;6:12-15.